Spatial analysis of room impulse responses captured with a
32‐capsules microphone array
A.Capra, A.Farina, A.Amendola and C.Varani andrea.capra@mac.com
Industrial Engineering Department University of Parma
Most of the acoustical measurements in theatres, concert halls and musical spaces are performed by using a single omnidirectional pressure microphone.
Introduction
Used for reverberation time and other monophonic parameters
NO information about the arrival direction of the sound
Aim of the research: visual and dynamic representation of the acoustical behaviour of the room under test (theatres, concert halls, etc.).
The dynamic vision of this behaviour could be useful for :
• Finding unwanted reflections (echoes)
• Evaluating the spectral content of those reflections
• Checking if the sound reinforcement loudspeakers are correctly located and aimed
Having at disposal directive microphones and aiming
them in different directions it should be possible to obtain
a chart of the behaviour of the sound in time and in
frequency
First attempt: a rotating shotgun microphone
2008 - two similar opera houses chosen for the measurements:
Teatro Sociale
di Como Teatro Comunale di
Modena
Measurements equipment:
- Omnidirectional source
- Edirol FA-101 sound card and a laptop for the recording of the sine-sweep test signal
- Sennheiser ME66 directive microphone mounted on a
Outline rotating table
- Azimuth: 18 steps (20°) - Elevation: 8 steps (22.5°)
- Impulse responses derived from sweeps
by using Adobe Audition 1.5 and Aurora
plug-ins
Some results
Teatro Comunale di Modena Source in orchestral pit
Point A
125 Hz 1000 Hz 4000 Hz
24 ms (direct sound)
The direct sound is rich of low frequencies for the diffraction of the pit
40 ms (first reflection)
The high frequencies arrive to the performer 16 ms after the direct sound
Teatro Sociale di Como
S1 R
Source S1: the first reflection arrives from the back wall of the theatre after 40 ms
S2
4000 Hz
Source S2: direct sound at high frequencies is weaker than the first reflection coming after 56 ms from the proscenium arch
4000 Hz
Dynamic polar plot in vertical and horizontal plane.
Teatro Sociale di Como – 4000 Hz
Reflection from the proscenium arch
The new approach
How can we obtain lots of directive microphones with only one probe? Using an array of capsules!
2009: first prototype of a spherical array (32 capsules)
Expanded polyurethane (too much delicate for
handling!)
32 capsules for earing-aid
(poor quality)
Eigenmike
®by mhAcoustics
• 32 high quality capsules
• Pre-amplifiers and A/D converters packed inside the sphere
• All the signals are delivered to the audio interface through a digital CAT-6 cable
• The audio interface is an EMIB Firewire interface based on TCAT DICE chip:
- supported by any OS
- 8 digital output(ADAT) + 2 analogue output - Wordclock
• The pre-amplifier gain control is operated by MIDI control
…in the meantime …
The signal processing
The idea Synthesis of 32 directive virtual microphones in the direction of the capsules employing a set of digital filters
M = 32 signals coming from the capsules
V = 32 signals yielding the desired virtual microphones
Bank of MxV FIR filters
y
v(t) = x
m(t) ∗ h
m.v(t)
m=1
∑
MOutput signal of V mic.
Input signal from the m-capsule
Matrix of filters
Traditional design of the filters
The processing filters hmv are usually computed following one of several, complex mathematical theories, based on the solution of the wave
equation (often under certain simplifications), and assuming that the microphones are ideal and identical
In some implementations, the signal of each microphone is processed through a digital filter for compensating its deviation, at the expense of heavier computational load
outputs of the microphone array are maximally close to the prescribed ideal
responses This method also inherently corrects
for transducer deviations and acoustical artefacts (shielding, diffractions, reflections, etc.)
No theory is assumed: the set of hmv filters is derived directly from a set of impulse response measurements, designed according to a least- squares principle.
Novel approach
Virtual microphones synthesized for this research:
4th ORDER CARDIOIDS Matlab script
•Inputs:
9 2048 samples of each IR
9 The number of virtual microphones 9 Directivity of each virtual microphone 9 Azimuth and elevation of each virtual
microphone
IRs Matrix inversion
Output: FIR filters matrix
✕ =
The new session of measurements Two different kind of musical space:
Sala dei Concerti
(Casa della Musica - Parma) Teatro “La Scala”
(Milano)
Equipment:
•Eigenmike
®probe as receiver
•Laptop as recorder
•“Lookline” Dodecahedron as omni-direcitonal source
•Sine sweep as test signal
The post processing software
MATLAB script
Polar plot of the sound levels in the horizontal planeMesh map of the levels on a 360°x180° picture of the room
Dynamic plots
1
ststep
Import of WAV files containing the IRs obtained from the
32 virtual microphones
Frequency bands of analysis Impulse Response length
Pressure Level (dB) Root 2
Root 3
2
ndstep
FFT processing
The results
The probe is not calibrated with an absolute level: every point of measure has its own level normalization and
colour scale.
NEVERLES
Growing the distance between source and receiver the S lobe of the reflections becomes more relevant in
comparison with the direct sound.
1 3 2
24
24 virtual microphones for horizontal polar
1
2 32 3
32 virtual microphones for 3D map
Teatro “La Scala”: 4 kHz – root 2 – 8000 samples
The direct sound
500 Hz
1000 Hz Radiation of the
wooden stage
The source is close to the receiver
The direct sound masks on the map the first reflection on the floor
The first visible reflection comes from the column behind the source, followed by a reflection on the floor and by a reflection coming from the opposite direction of the first one.
The reflections
4 ms
57 ms
The sound bounces
between the columns of the proscenium arch…
… and then arrives the ceiling reflection.
Some notes
• The shape of the theatre refocuses the sound in the point of provenience
• Lobes in the 500 Hz band: they are effect of the diffraction
• The reflections coming from the walls are displayed
with cold colours: big amount of absorption
“Sala dei Concerti”: 4 kHz – root2 – 8000 samples
The direct sound
500 Hz - the sound in this freq. band anticipates the direct sound. Is the effect of the radiation of the wooden structure of the stage
500 Hz
4000 Hz
The first reflection from the floor is clearly visible, followed by the diffuse sound of the wooden diffusors.
21 ms
Strong lateral reflection (90°)
coming from a plane even surface.
On the opposite side that reflection is not
present because of an
absorbent curtain.
After this reflection the sound appears quite diffuse in the room apart from a little bounce between the lateral walls.
97 ms
Strong reflection from the back of the room.
The effect is audible
also from the stage and
causes problems with
high repeated notes to
the performers
Conclusions
Today we presented a new approach to the impulse response measurements. It permits the dynamic view of the impulse responses plotted on a panoramic picture (360°x180°) by using 32 high directive virtual
microphones.
Positive aspects
Easy visualization of undesired reflections
Easy way for locating their arrival direction
Easier way for correcting them
This is a first step in the research, a lot of work has to be done:
• Calibration of the probe
• Creation of a plug-in (VST or Audacity-based)
• IP-camera for taking the panoramic pictures
Thank you for the attention
andrea.capra@mac.com University of Parma